Re: [rdiff-backup-users] keeping hard-linked copy while backup in progress

[Jakob Unterwurzacher]

Dave Kempe schrieb:
>
> Hi Michael,
> no, thats not quite correct - the increments are reverse differentials,
> and rdiff-backup handles each backup atomically, so if the backup is
> interrupted for any given backup run, then the changes in files that are
> relevant to that backup (the files that have changed or been added etc)
> are effectively lost on the destination. This is because rdiff-backup
> rolls the backup back to the last known good state on the next run if
> the previous run was interrupted - it calls this 'regresssing destination'.
> 
> The only time I have lost all the increments and have them become
> useless is when the filesystem on the destination is corrupted and files
>
> [...]

Probably things have changed since then, but on Ubuntu Dapper i tried to
backup over an unstable DSL connection (disconnect every 24h or so).
Every once in a while the repository became corrupted (i.e. rdiff-backup
crashes on every subsequent run until i blow away the increments). I
wouldn't recommend running rdiff-backup over an unstable link.

Coming backup to

> My question is, does rdiff-backup update any files in-place in the target 
> folder?

For files not in rdiff-backup-data: it doesn't. I just tested that. It
creates a new file and then moves  ("mv -f") it over the original file.
A hard link retains the original contents.
For files in rdiff-backup-data: i have no idea. You could try it out: cp
-al, create md5sums for everything, then rdiff-backup again, then check
the sums.

Jakob